8.1 Optimization Modeling: Decision Variables, Objectives, and Constraints

8.1 Optimization Modeling: Decision Variables, Objectives, and Constraints

Mathematical programming is the cornerstone of operations research and industrial systems engineering. In modern production facilities, logistics networks, and supply chains, engineers routinely face complex decisions regarding how to allocate scarce resources—such as machine hours, labor shifts, raw materials, capital, and warehouse space—to achieve an optimal operational objective. The NCEES FE Industrial and Systems examination tests your ability to translate verbal problem descriptions into clean, rigorous mathematical formulations and evaluate their underlying assumptions.

---

1. The Three Pillars of Mathematical Programming

Every mathematical optimization model, regardless of scale or complexity, is constructed upon three indispensable elements:

                    ┌────────────────────────────────────────┐
                    │         MATHEMATICAL PROGRAM           │
                    └───────────────────┬────────────────────┘
                                        │
         ┌──────────────────────────────┼──────────────────────────────┐
         ▼                              ▼                              ▼
┌─────────────────┐           ┌───────────────────┐          ┌────────────────────┐
│ 1. DECISION     │           │ 2. OBJECTIVE      │          │ 3. CONSTRAINTS     │
│    VARIABLES    │           │    FUNCTION       │          │                    │
│  What decisions │           │  Single criterion │          │  Resource limits,  │
│  must be made?  │           │  to max or min    │          │  demand, bounds    │
│  (x_1, x_2,...) │           │  Max/Min Z = c^T x│          │  Ax <= b, x >= 0   │
└─────────────────┘           └───────────────────┘          └────────────────────┘

1. Decision Variables ($x_j$)

Decision variables represent the quantifiable choices under the direct control of the decision-maker. Formulating an optimization model always begins by defining decision variables with precise units and explicit temporal horizons.

$\text{Example: } x_j = \text{number of units of product } j \text{ produced per week [units/week]}$

Exam Watchout: Never define a decision variable vaguely (e.g., "let $x_1 = \text{Product 1}$"). A decision variable must represent a numerical quantity with explicit units of measurement (e.g., "$x_1 = \text{gallons of chemical additive A blended per operating shift}$").

2. Objective Function ($Z$)

The objective function expresses the single operational performance measure to be optimized as a function of the decision variables. In linear programming, this function is strictly a linear combination of the decision variables:

$\text{Maximize or Minimize } Z = \sum_{j=1}^{n} c_j x_j = c_1 x_1 + c_2 x_2 + \dots + c_n x_n$

Where $c_j$ represents the objective function coefficient (e.g., unit profit contribution, unit revenue, unit acquisition cost, or unit cycle time) associated with activity $j$.

3. Constraints

Constraints represent technological limitations, physical capacities, legal regulations, or contractual requirements that restrict the values the decision variables can assume. They are categorized into:

1. Structural (Technological) Constraints: Functional restrictions imposed by limited resources: $\sum_{j=1}^{n} a_{ij} x_j \le b_i \quad \text{(resource capacity limitations)}$ $\sum_{j=1}^{n} a_{ij} x_j \ge b_i \quad \text{(minimum service levels or contractual requirements)}$ $\sum_{j=1}^{n} a_{ij} x_j = b_i \quad \text{(exact physical blending or balance requirements)}$ Where $a_{ij}$ is the technological consumption coefficient (units of resource $i$ consumed per unit of activity $j$), and $b_i$ is the right-hand-side (RHS) available capacity or requirement.

2. Sign Restrictions (Non-Negativity): $x_j \ge 0 \quad \forall j = 1, 2, \dots, n$ Negative production or negative inventory transfers are physically meaningless in standard industrial settings. If a variable is allowed to take positive, zero, or negative values (e.g., net cash flow or coordinate deviations), it must be explicitly defined as Unrestricted in Sign (URS) and represented as the difference between two non-negative variables: $x_j = x_j^+ - x_j^-$, where $x_j^+, x_j^- \ge 0$.

---

2. Fundamental Assumptions of Linear Programming

For a mathematical model to qualify as a valid Linear Program (LP), four core mathematical assumptions must hold simultaneously:

                     ┌───────────────────────────────────┐
                     │       LP CORE ASSUMPTIONS         │
                     └─────────────────┬─────────────────┘
           ┌─────────────────┬─────────┴─────────┬─────────────────┐
           ▼                 ▼                   ▼                 ▼
   ┌───────────────┐ ┌───────────────┐   ┌───────────────┐ ┌───────────────┐
   │PROPORTIONALITY│ │  ADDITIVITY   │   │  DIVISIBILITY │ │   CERTAINTY   │
   │  c_j is fixed │ │ No cross-terms│   │  Continuous   │ │ Deterministic │
   │  regardless   │ │ (no x_1*x_2)  │   │  fractional   │ │  parameters   │
   │  of scale     │ │ Sum of parts  │   │  values valid │ │  (no randoms) │
   └───────────────┘ └───────────────┘   └───────────────┘ └───────────────┘

---

3. Standard Formulation Taxonomy & Problem Archetypes

The FE Industrial exam frequently presents optimization scenarios drawn from four classic problem classes. Recognizing these patterns allows you to establish the correct decision variables and constraints in seconds.

1. Product Mix Models

A facility manufactures $n$ distinct products using $m$ shared, limited resources (machine capacities, skilled labor hours, raw material inventories). The goal is to determine the production volume of each product to maximize total contribution margin.

\text{Maximize } & Z = \sum_{j=1}^{n} c_j x_j \\ \text{Subject to: } & \sum_{j=1}^{n} a_{ij} x_j \le b_i, \quad \forall i = 1, \dots, m \\ & x_j \ge 0, \quad \forall j = 1, \dots, n \end{aligned}$$ ### 2. Blending and Quality Specification Models Blending problems require determining the optimal proportions of raw feedstocks (e.g., crude oils, chemical reagents, metal alloys, animal feeds) to produce an end product that meets strict physical, chemical, or nutritional property thresholds at minimum total cost. A common challenge on the FE exam is **linearizing ratio and percentage constraints**. Suppose feedstock $j$ has a property rating $q_j$ (such as octane number or sulfur percentage) and costs $c_j$ per gallon. The blended fuel must have a minimum weighted property rating of $Q_{\min}$: $$\frac{\sum_{j=1}^{n} q_j x_j}{\sum_{j=1}^{n} x_j} \ge Q_{\min}$$ As written, this constraint is **non-linear** because decision variables appear in both the numerator and denominator. Because $\sum x_j > 0$, we cross-multiply the denominator and collect like terms to produce a strictly linear constraint: $$\sum_{j=1}^{n} q_j x_j \ge Q_{\min} \sum_{j=1}^{n} x_j \implies \sum_{j=1}^{n} (q_j - Q_{\min}) x_j \ge 0$$ ### 3. Transportation and Transshipment Networks In a classical **Transportation Problem**, a commodity must be shipped from $m$ supply origins (factories, warehouses) with capacities $s_i$ to $n$ demand destinations (retailers, regional depots) with requirements $d_j$ at minimum total transit cost. Decision variables are defined as: $$x_{ij} = \text{units shipped from origin } i \text{ to destination } j$$ $$\begin{aligned} \text{Minimize } & Z = \sum_{i=1}^{m} \sum_{j=1}^{n} c_{ij} x_{ij} \\ \text{Subject to: } & \sum_{j=1}^{n} x_{ij} \le s_i, \quad \forall i = 1, \dots, m \quad \text{(Supply constraints)} \\ & \sum_{i=1}^{m} x_{ij} \ge d_j, \quad \forall j = 1, \dots, n \quad \text{(Demand constraints)} \\ & x_{ij} \ge 0, \quad \forall i, j \end{aligned}$$ In a **Transshipment Problem**, intermediate transshipment nodes (cross-docks, consolidation centers) exist where goods enter and leave. For every intermediate node $k$, a **flow conservation constraint** must hold: $$\sum_{\text{all } i} x_{ik} - \sum_{\text{all } j} x_{kj} = 0 \quad \iff \quad \text{Total Inflow} - \text{Total Outflow} = 0$$ ### 4. Production Planning and Multi-Period Scheduling Models Manufacturing systems must plan production over a discrete planning horizon of $T$ time periods (weeks or months) with fluctuating seasonal demand $D_t$. In each period $t$, the system can produce units using regular time ($P_t$), overtime ($O_t$), or subcontracting ($S_t$), and carry unsold units into future periods as inventory ($I_t$). The fundamental governing relationship is the **Inventory Balance Equation**, which enforces physical continuity across successive time periods: $$I_{t} = I_{t-1} + P_t + O_t + S_t - D_t, \quad \forall t = 1, \dots, T$$ Rearranging to standard linear form with decision variables on the left and known parameters on the right: $$I_{t-1} + P_t + O_t + S_t - I_t = D_t$$ Where $I_0$ is the known initial inventory at the start of the planning horizon. --- ## 4. Step-by-Step Methodology for Converting Word Problems into LP Formulations When confronting a complex verbal optimization question on the FE exam, follow this systematic five-step protocol: 1. **Step 1: Identify the Decisions and Define Variables** Ask: *"What numerical decisions must be made to solve this problem?"* Write down each variable $x_j$ with explicit units (e.g., units/hour, kg/batch, dollars invested). 2. **Step 2: Formulate the Objective Function** Ask: *"What is the overarching operational goal?"* Determine whether the goal is maximization (profit, revenue, efficiency, yield) or minimization (cost, scrap, cycle time, transit distance). Express $Z$ as a linear combination of the decision variables. 3. **Step 3: Tabulate Resource Limits and Draft Structural Constraints** Identify every limiting factor mentioned in the prompt. Group them into logical categories: capacity limits ($\le$), minimum production/demand targets ($\ge$), and exact balance/conservation equations ($=$). Ensure that each constraint has variable terms on the LHS and pure constants on the RHS. 4. **Step 4: State Non-Negativity and Variable Bounds** Explicitly specify $x_j \ge 0$ for all variables, or indicate any unrestricted/bounded variables ($L_j \le x_j \le U_j$). 5. **Step 5: Perform Dimensional Consistency Analysis** Verify that the units of measurement on both sides of every constraint match exactly. If the RHS represents hours, the LHS must sum to hours (e.g., $\text{hours/unit} \times \text{units} = \text{hours}$). If cycle times are given in minutes, convert either the LHS or the RHS so that time bases are uniform. --- ## 5. Step-by-Step Worked Engineering Examples ### Worked Example 8.1.1: Product Mix in an Advanced Electronics Facility **Problem**: An industrial electronics plant manufactures two types of programmable logic controllers: Standard ($x_1$) and Industrial ($x_2$). Each controller must pass through three sequential processing centers: Surface Mount Technology (SMT) Placement, Automated Optical Inspection (AOI), and Environmental Stress Testing. The processing times per unit, total available weekly machine capacities, and net unit profit contributions are summarized below: | Work Center | Standard ($x_1$) [hours/unit] | Industrial ($x_2$) [hours/unit] | Total Available Capacity [hours/week] | | :--- | :---: | :---: | :---: | | SMT Placement | 1.5 | 2.5 | 120 | | AOI Inspection | 0.8 | 1.2 | 64 | | Environmental Testing | 2.0 | 1.0 | 90 | | **Unit Profit Contribution** | **$75** | **$110** | — | In addition, a corporate supply agreement mandates that the facility must deliver at least 20 total controllers per week to keep its prime vendor certification, and customer demand for the Industrial model is capped at 40 units per week. Formulate the complete linear program. **Solution**: 1. **Define Decision Variables**: - $x_1 = \text{number of Standard controllers assembled per week [units/week]}$ - $x_2 = \text{number of Industrial controllers assembled per week [units/week]}$ 2. **Formulate the Objective Function**: Maximize total weekly profit contribution: $$\text{Maximize } Z = 75 x_1 + 110 x_2$$ 3. **Formulate Structural Constraints**: - SMT Placement capacity limit: $$1.5 x_1 + 2.5 x_2 \le 120$$ - AOI Inspection capacity limit: $$0.8 x_1 + 1.2 x_2 \le 64$$ - Environmental Stress Testing capacity limit: $$2.0 x_1 + 1.0 x_2 \le 90$$ - Minimum total production commitment: $$x_1 + x_2 \ge 20$$ - Maximum demand ceiling for Industrial model: $$x_2 \le 40$$ 4. **State Non-Negativity Constraints**: $$x_1 \ge 0, \quad x_2 \ge 0$$ ### Worked Example 8.1.2: Specialty Chemical Blending with Linearized Octane Specification **Problem**: A refinery produces high-grade aviation fuel by blending two intermediate petroleum feedstocks: Alkylate ($x_A$) and Reformate ($x_R$). The chemical properties and unit costs per barrel are as follows: | Feedstock | Research Octane Number (RON) | Vapor Pressure (psi) | Cost per Barrel ($/bbl) | | :--- | :---: | :---: | :---: | | Alkylate ($x_A$) | 98 | 4.5 | $92 | | Reformate ($x_R$) | 92 | 2.0 | $80 | The final blend must satisfy three quality standards: 1. Total production must be exactly 10,000 barrels. 2. The blended Research Octane Number (RON) must be at least 95. 3. The blended vapor pressure must not exceed 3.5 psi. Formulate the linear programming model to minimize total blending cost. **Solution**: 1. **Decision Variables**: - $x_A = \text{barrels of Alkylate blended [bbl]}$ - $x_R = \text{barrels of Reformate blended [bbl]}$ 2. **Objective Function**: Minimize total cost: $$\text{Minimize } Z = 92 x_A + 80 x_R$$ 3. **Structural Constraints**: - Total volume requirement: $$x_A + x_R = 10{,}000$$ - Minimum Octane Specification (Linearization): $$\frac{98 x_A + 92 x_R}{x_A + x_R} \ge 95$$ Multiply across by $(x_A + x_R)$ (valid since $x_A + x_R = 10{,}000 > 0$): $$98 x_A + 92 x_R \ge 95(x_A + x_R)$$ $$98 x_A + 92 x_R \ge 95 x_A + 95 x_R$$ Collect terms on the left: $$(98 - 95) x_A + (92 - 95) x_R \ge 0 \implies 3 x_A - 3 x_R \ge 0$$ - Maximum Vapor Pressure Specification (Linearization): $$\frac{4.5 x_A + 2.0 x_R}{x_A + x_R} \le 3.5$$ $$4.5 x_A + 2.0 x_R \le 3.5(x_A + x_R) \implies (4.5 - 3.5) x_A + (2.0 - 3.5) x_R \le 0$$ $$1.0 x_A - 1.5 x_R \le 0$$ 4. **Non-Negativity**: $$x_A \ge 0, \quad x_R \ge 0$$ --- ## 6. NCEES Reference Handbook Tips & Realistic Exam Traps - **The Unit Conversion Trap (Minutes vs. Hours)**: One of the most frequent errors on the FE exam occurs when task times are given in minutes (e.g., $15\text{ min/unit}$) while available work center capacity is given in hours (e.g., $40\text{ hours/week}$). You must convert either minutes to hours ($15/60 = 0.25\text{ hr/unit}$) or hours to minutes ($40 \times 60 = 2400\text{ min}$). Mixing $15 x_1 + 25 x_2 \le 40$ creates an immediate factor-of-60 error. - **The Fractional Constraint Trap**: Never submit an optimization model with variables in the denominator. A constraint like $\frac{x_1}{x_1 + x_2} \ge 0.40$ is technically non-linear in that format. Always cross-multiply to form $0.6 x_1 - 0.4 x_2 \ge 0$. - **The Direction of Optimization in Cost vs. Profit**: Double check whether the prompt specifies maximizing profit or minimizing cost. Selecting the wrong optimization sense will completely invert your solution. - **Sign Restrictions for Net Balances**: In cash-flow or temperature deviation models where variables can drop below zero, do not automatically append $x_j \ge 0$. Note whether a variable is unrestricted in sign (URS).